Researchers show that simple measurements of how well blood serum neutralizes viruses can tell us about the internal makeup of an antibody response. Using high-throughput neutralization data against 78 influenza H3N2 strains, the team found that the pattern of measured titers across people can indicate whether protection comes from many antibodies acting together or from a few very strong antibodies.

The group reanalyzed a dataset produced with a barcoded viral library and a sequencing-based neutralization assay. In that experiment, each viral variant carried a unique barcode. Sera were tested at serial dilutions from about 40-fold to 10,000-fold. The neutralization titer T was defined as the dilution that cuts the virus’s infectivity by half. The undiluted serum concentration used in the original work was c0 = 667 nM. To compare different viruses fairly, the authors rescaled titers for each virus and then worked with the transformed quantity F = −ln T.

The rescaled data revealed clear differences between groups. Post-vaccination adults showed a roughly Gaussian (bell-shaped) distribution of F values, while sera from children were asymmetric and far from Gaussian. Quantitatively, the adult mean of F was about −6.0 with a standard deviation of 0.9, and the child mean was about −5.0 with standard deviation 1.0. The child data had pronounced negative skew (skewness ≈ −0.80) with a p-value near 10^−70 for departure from zero skewness. Titers averaged across viruses were also strongly correlated between the two groups (Pearson r = 0.75).

To explain these patterns the authors developed a simple equilibrium binding model. They represent a serum as a mix of N distinct antibody clonotypes (each clonotype is a family of related antibodies from one B‑cell lineage) at concentrations fa. Each clonotype is assumed to target one viral site, or epitope, and to have a dissociation constant KD that measures binding strength (smaller KD means stronger binding). They assume a virus is neutralized if any antibody binds it. In the simple one-epitope limit, the titer T equals c0/KD, so F = ln(KD/c0). More generally, F can be written as a function of the mix of clonotype concentrations and affinities. Under this view, a cohort can be “collective” (many moderate antibodies give Gaussian-like variation) or “extreme” (a few very strong antibodies dominate and produce a broad, asymmetric distribution). When extremes dominate, the statistics follow a Gumbel distribution from extreme value theory, which describes maxima drawn from many random variables.